Many existing fairness criteria for machine learning involve equalizing some metric across protected groups such as race or gender. However, practitioners trying to audit enforce group-based can easily face the problem of noisy biased group information. First, we study consequences naively relying on labels: provide an upper bound violations true G when are satisfied $\hat{G}$. Second, introduce two new approaches using robust optimization that, unlike naive approach only $\hat{G}$,...

Recommender systems play a crucial role in mediating our access to online information. We show that such algorithms induce particular kind of stereotyping: if preferences for set items are anti-correlated the general user population, then those may not be recommended together user, regardless user's and rating history. First, we introduce notion joint accessibility, which measures extent can jointly accessed by users. study accessibility under standard factorization-based collaborative...

Directed network with flow dynamics is an important topic in realistic complex systems such as banking systems. We study the stability of these financial networks using a homogeneous and more general inhomogeneous model. The nodes are banks they under four different kinds disturbance that may lead to bankruptcy, which defined by condition capital value bank below critical value, zero. this measured minimum after fixed time money network. smaller higher probability first bankruptcy. By means...

We provide theoretical complexity analysis for new algorithms to compute the optimal transport (OT) distance between two discrete probability distributions, and demonstrate their favorable practical performance over state-of-art primal-dual capability in solving other problems large-scale, such as Wasserstein barycenter problem multiple distributions. First, we introduce \emph{accelerated randomized coordinate descent} (APDRCD) algorithm computing OT distance. its upper bound...

While many areas of machine learning have benefited from the increasing availability large and varied datasets, benefit to causal inference has been limited given strong assumptions needed ensure identifiability effects; these are often not satisfied in real-world datasets. For example, observational datasets (e.g., case-control studies epidemiology, click-through data recommender systems) suffer selection bias on outcome, which makes average treatment effect (ATE) unidentifiable. We propose...

We study the problem of learning revenue-optimal multi-bidder auctions from samples when bidders' valuations can be adversarially corrupted or drawn distributions that are perturbed. First, we prove tight upper bounds on revenue obtain with a distribution under population model, for both regular valuation and monotone hazard rate (MHR). then propose new algorithms that, given only an ``approximate distribution'' bidder's valuation, learn mechanism whose is nearly optimal simultaneously all...

Abstract This paper presents two public-key cryptosystems based on the so-called expanded Gabidulin codes, which are constructed by expanding codes over base field. Exploiting fast decoder of we propose an efficient algorithm to decode these new when noise vector satisfies a certain condition. Additionally, have excellent error-correcting capability because optimality their parent codes. Based different masking techniques, give encryption schemes using in McEliece setting. According our...

We introduce the "inverse bandit" problem of estimating rewards a multi-armed bandit instance from observing learning process low-regret demonstrator. Existing approaches to related inverse reinforcement assume execution an optimal policy, and thereby suffer identifiability issue. In contrast, we propose leverage demonstrator's behavior en route optimality, in particular, exploration phase, for reward estimation. begin by establishing general information-theoretic lower bound under this...

This paper examines strategies of asymmetric enterprises in joint pollutant abatement with collective-risk. Asymmetric are divided into two types: one type high initial resource endowments, the other low endowments. Evolutionary dynamics distinct scenarios namely fair contribution and altruistic preference analyzed. The proves that endowments have greater incentives to cooperate, more likely be agents. Moreover, effects increment risk rate on cooperation sided. A large group size hinders...

We study the problem of finding equilibrium strategies in multi-agent games with incomplete payoff information, where matrices are only known to players up some bounded uncertainty sets. In such games, an ex-post characterizes that robust uncertainty. When game is one-shot, we show zero-sum polymatrix can be computed efficiently using linear programming. further extend notion stochastic played repeatedly a sequence stages and transition dynamics governed by Markov decision process (MDP)....

Abstract This paper presents a new technique for disturbing the algebraic structure of linear codes in code-based cryptography. Specifically, we introduce so-called semilinear transformations coding theory and then apply them to construction cryptosystems. Note that Fqm can be viewed as an Fq -linear space dimension m , transformation φ is therefore defined automorphism . Then impose this code C over It clear (C) forms space, but generally does not preserve -linearity any longer. Inspired by...